#tc-research
1 messages ยท Page 11 of 1
@proven dawn There was no fight, two simulations came up with different results, I wanted to compare the two. This isn't a Discord for simc or for AMR. I think it's reasonable to ask questions regarding different APL's.
Yeah thats fine, digging into the two APLs to figure out why they are producing different results is entirely fair and valid
Hi, not sure if this is the correct channel but the latest SimC (latest OS X nightly, to be specific) is giving mastery a small negative stat weight when I sim as Outlaw. (-0.01 with 25k iterations) Known issue?
yup, its on github, known issue
Sorry if discussed already but if there are some issues with KB the numbers i saw in raid after the patch are not the numbers blizz wana see for AP or EX Sin? Feels like I have to prepare for a nerf after the nerf?
@cosmic orbit @split sorrel by request, I added a simulation with shoulder and bracers to that mmo-c thread I made. Still not getting any gains by re-rolling a 2 buff combo in the AMR simulator. I also tried re-created the exact SimC rotation in AMR and comparing the two simulators. I don't see any bugs in either, but still never get a gain by ditching two buffs, even using SimC rotation.
The two simulators work slightly differently, though, so some small differences are to be expected
Also while I'm here, I figure I should drop our guide for running weights on AMR to join the ranks of the other great guides for simc.
A picture tutorial to run stat weights: http://i.imgur.com/r2gpuBW.jpg
(that's all, not here to promo amr, just wanted to provide a quick guide)
@reef canyon one thing that keeps me from using AMR often is simulationtime, running a low error % gearing sim with machine learning seems to take upwards of 4-8 hrs on my computer, whereas a similar one in SimC takes about 30 minutes
any plans to improve simulation speed?
@maiden cargo the SimC scaling and AMR gearing strategy do two very different things, that's why AMR takes so long. SimC generates a couple of data points for each stat and gives you some local scaling for stats. AMR generates 1000 (or however many you choose) data points for random stats across the board and analyses those. Then it simulates all of the trinkets, legendaries, and set bonuses and caches that info to use with the gear optimizer as well. Lowing the error % below 0.25% in AMR doesn't actually give you better predictive power, so you don't really need to do that - that will cut the time down substantially. 500 data points at 0.5% margin of error is really going to give you almost as good data as 1000 points at 0.25%.
you can also toggle "plots" in simc and get a million data points
Yeah, you can script SimC to make whatever data points you want, and then analyze them however you want. We just "canned" a method for ranking as much gear as possible, since that is what we want to do on our website.
additionally, is there any particular reason why the AMR rerolling strategy of RTB is so bad?
specifically it seems to be keeping any 2+ buff combination
and rerolling all singles
I just made a post on that today actually
Hey all, it's been a while since I made a theorycraft post in the rogue forums. I've seen some threads here and there suggesting new logic for what the best buffs to keep on RtB are. A lot of folks are suggesting that you really want to roll until you get True Bearing, unless you get a really lucky 3 or 6 buff combo. So, I decided to gather a ton of data points to help answer the question.
First, I used an 880 ilvl rogue with no specific legendaries, set bonuses, or "special" trinkets. I
short answer is that it's still simulating to the best strategy, in all the tests I've done.
although keeping True Bearing seems to be about equal now, so I'm going to change it to keep True Bearing as a single buff
IMO, the more options the better, reduces variance
ill read this all the way through it looks pretty interesting
but i did spot a particular issue "I decided that the outlaw legendaries wouldn't have a significant effect on this particular topic"
i ended up doing a test on some in a subsequent post
they didn't end up changing the results
but i can run another large test on them as well at some point
as far as i know, mantle and greenskins cuffs both increase the value of true bearing quite a bit
the boots have an impact aswell
An analysis of the difference between the SimC and AMR results on rerolling w/r/t legendaries would be interesting
even ignoring legendaries would be a very interesting study if someone wants to write it up
That is the value of having multiple tools, cross validation, if we aren't doing that whats the point in having multiple tools
I've been doing a lot of comparisons between the two simulators since 7.1.5. I'll see if I can come up with a nice comparison sometime soon. The difficult thing is that the nuances of how the two simulators work start to come into play when you're looking at small differences - especially differences that relate to timing. The two simulators advance time differently and use a different model for choosing the next action. I have been working on a write-up to try to explain the differences to theorycrafters so they can get the most out of comparisons between the two tools. Sometimes people really want to say one or the other is "wrong" or "right"... when really they are two projects that have made different choices on how to model the game.
Well in theory one could be more right or wrong since we are comparing against a known truth in the form of in-game but for now we need to understand the differences
Yeah, when we made the AMR simulator from the ground up for Legion, we built in what we call "anticipation" logic to mimic how a player needs to queue up actions in the game. There are situations in SimC that allow you to act upon information that wouldn't actually be available to you yet when you choose your action. You can get around some of it by using ".react" in SimC APLs, but, as far as I know, you can't set up a way to make that work with Power (Energy) in SimC.
I'll let @split sorrel or someone else who writes the SimC APLs handle that one because I'm not an expert on all the intricacies of SimC APLs
@manic pulsar how does one edit/customize the APL/rotation on AMR?
Especially when you are in a situation where you are waiting for Energy, lets say you are waiting for 50 Energy to use Saber Slash - SimC will use SS when you have exactly 50 energy every time. Players just can't do that perfectly. In AMR you will use SS when you have close to 50 energy, like within 1-3 energy of 50. That, imo, is more realistic - and little stuff like that can add up over the iterations.
@round canyon you can go to the http://www.askmrrobot.com/wow/simulator/run page and then load up the player you want to sim. Pick the rotation from the picker, then there is a view/edit link to the right of it. You can also see all the rotation on the wiki: http://www.askmrrobot.com/wow/theory/?spec=RogueOutlaw&version=live. When viewing a rotation, click the Copy button to copy it and then an edit option becomes available. Your custom ones are saved with your username.
so i assume that the reason the edit/view link doesnt allow me to edit anything is because i dont have an account?
AMR just uses my curse account, correct?
no
i was under the impression curse owned amr but i guess they must not then
we have no access to curse data
we are an independent company
just 3 people
but we use wowdb for item data, that's all
this is the SimC APL translated to AMR:
it looks a lot more verbose because we don't support implicit type conversion
the MaxCombo parameter can be found in the Parameters section that isn't expanded by default
why do i sim like 30k more dps with the new build, exact new settings like i always had
probably kingsbane bug implemented
kingsbane has been stacking 30% per stack with AP for quite a while. looks like aethys updated it yesterday in SimC
i think it's been like that since 7.1 at least
Ok, but its correct now?
i mean, when i sim my character.. does it show the correct damage now?
sorry for being retarded
in SimC? I'm not sure, haven't looked today. It does in AMR
well, i wont complain if its correct i mean 30k dps more.. and i just got 4 set aswell
so simulating will be fun D:
@split sorrel what did you do with kingsbane, is my sims correct now? or is something wrong
@manic pulsar i noticed your ghostly strike logic seemed a little wonky, it was casting ghostly strike during curse or casting it while at max combo points pretty frequently
i changed it to the following and noticed a 4-5k dps increase in the simulation
BuffRemainingSec(GhostlyStrike) < 0.3*15 and AlternatePower < MaxAlternatePower or (CooldownSecRemaining(CurseOfTheDreadblades) < 3 and BuffRemainingSec(GhostlyStrike)<12)
rog_asn_t19m_1t_aa-4-3x-30-10_relics.html is this one for agonizing relics?
@round canyon cool I can update that. Hadn't really optimized that logic yet.
@split sorrel does the nightly build APL optimized for draught of souls?
i mean i think a few build before my sim dps with draught is lower than urn. then kiwifruit give me a custom APL to work with draught and it increases my dps by ~20k
actions.cds+=/use_item,slot=trinket1,if=cooldown.shadow_dance.charges_fractional<=2.45&buff.shadow_dance.down
this was it
there is even a better one with sprint according to what @torn dome said in #tc-updates
but I can't really work on things atm
oh didnt noticed that
are the trinket changes in the latest nightly of simcraft up to date?
not awful, just less better
@split sorrel Do you have a comparitive list of RtB buff combinations for DPS? Like how far under True Bearing or 3+ would Shark + anything be?
@stiff forum what draught combo?
draught into sprint vanish
you can use sprint/drought together bc. drought does not cancle the spring trait (for sub)
use at low energy and get a flippin free vanish
@Orckefeller#1827 i have a comparison i actually ran today using my character(not set bonuses)
let me find it for you
Oh nice!
woulnt this be even more ridicolous with mantle of the master assassin?
well mantle is just op af
Where's TB alone or Shark + anything?
I'm trying to decide if it's worth keeping Shark with CD's, on pull, etc.
Because purely rolling for TB or 3+ is not fun.
i didnt do those
and im honostly not sure how to change the APL to have that happen
Yeah, I don't know how to do it either.
our 2nd rogue got the drought of souls and it's already doing 7-10% of his damage as assassination, wouln't this be crazy as sub with mantle+sprint and using draught right before the vanish triggers?
not sure if everything is well explained/detailed, that's what I used recently
it doesn't include every combos, but you can add them easily
don't have time to do more sry ๐
Thanks so much, Aethys.
Just trying to figure out what I should be doing and in-game testing with outlaw isn't reliable lol.
do you have the 4p and do you have the shoulders?
without more information its hard to say why simc is giving you those numbers
but i would bet that those numbers are indeed correct
what traits are on your relics?
FT/BP/FT
yeah i can say for sure why your getting those numbers now
your haste is somewhat low and without the 4p or energy reduction relics the value of haste goes up a lot
with 2 fortune's strike relics the value of mastery goes up a fair amount
and outlaws stats are generally somewhat close to eachother already which makes them switching places in the priority pretty likely
for example heres my current weights
https://gyazo.com/5d6f3a76a87ead390da2ebb7ebec1af1
very close
The real answer is, unless something looks dramatically off in the SimC results in terms of attacks rates or attack damage you should follow SimC
should i sub out my Hunger of the Pack for the BTI 835 i just grabbed?
There might be an issue with Mastery tho
the mastery issue is specific to blade flurry situations
and causes the value spat out by simc to be lower than it actually is
If you ask me the real reason for this close weights is simply the low gear
It doesnt matter what you get at that level
Its the same for other specs on other classes too
low gear?
my stat weights are at 892 ilevel
I thoguht you were talking about the guy in blue gear
both actually
after 7.1.5 outlaw has insanely close stat weights compared to sin and compared to a lot of other classes
This has at least been the case for my alts
also certain things adjust stat values that are independent of ilevel
set bonuses and artifact traits being the 2 large things
@split sorrel i got DoS yesterday didn't see it in your sim lists for 1t sub
have you played around with it? would it be higher priority than a 2 capped dance?
@round canyon I was out for most of the day and just got back. I see you were asking about rotations (AMR). I have a video for an overview https://www.youtube.com/watch?v=6MVBF44vXvM and an older video that shows some of the RtB tests we did in July (which is still relevant, just change the logic) https://www.youtube.com/watch?v=NF2Nz94AXOA
@split sorrel what i noticed is that without the custom APL i showed u just now, simc doesnt use draught on cd. and it also used it in the middle of shd. that's the reason why default APL sims draught pretty low
Is Simulation Craft updated for tier sets?
Is the sample sequence 100% optimized for sin?
the sample sequence shows how the sim plays the rotation.
the apl isnt 100% optimized (it may never be 100%).
depends what's ur reference
if it's average player, it's probably above 100%, if it's top player maybe not
Understandable. Thanks
Anybody know if the latest nightly has the trinket changes?
Anyone noticed that assa. Leg boots says 86 energy spend instead of 65 on AMR?
Someone know a way to run Sims from mobile? Askmrrobot website is not loadong properly in my phone
the bag of poison got approximately 1+haste procs per minute if you look at wowhead. Is that a valid Information. and does simc and friends account for it?
just go home and do it
Uhu ty for the obvius answer i guess...
....sorry there is no secret hidden answer you need a special handshake to unlock
Nani, remote desktop?
A while back someone had a site as well that ran on his server and took input from armory, like no way to change apls etc.. but I can't recall
@radiant dust
I just tried to run it right now, it worked
'estimated run time 1 hour' what
:o I don't agree much with amr APL
Well well
Is Simcraft simulating DoS correctly? I got a big dps loss, when i sim with it
As an assa
Yes but, why do you think it's modelled incorrectly? Do you get vastly different tick sizes, times, targets etc in simc compared to live?
Well, i just thought it should be better.. everyone is in a hype for the trinket ^^
Or is it more the usage of it in the APL? In that case, what is it sacrificing to use it? Does it cap energy due to it or cause something else to be delayed?
(Is it even using it? How did you generate the APL? etc)
Well, to be honest... i have absolutely no clue :/
Can you send me your APL on pastebin?
@strong gorge what don't you agree with in the AMR APL? I haven't done much work on it since 7.1.5 yet - was going to add in the sprint-vanish thing and draught of souls today for fun, since folks are talking about it. And, the simulation time will only be long if you choose "gearing strategy" as the simulation type - that is really going to do about 2500 simulations of varying margine of erro to create a complete gearing strategy for use with the gear optimizer. You can do smaller batches of sims or single runs if you don't want all that extra data.
@manic pulsar Sorry can't answer right now, in a meeting and then I'm leaving work. I'll try to answer tonight
does simcraft take the sin bracer effect into account?
yes
hey guys, i got these numbers, these scale deltas/error looks a bit weird, is this normal?
( Pawn: v1: "Piratenkarl": Agility=19.04, Ap=18.19, CritRating=13.98, HasteRating=8.94, MasteryRating=16.95, Versatility=12.89, Dps=9.60 )
is mine
its really high aswell
my stat weights for outlaw are much different. What gear, legendaries, and trinkets will definitely change your weights a lot
shoulders/2pc/4pc are the big stat weight changers
@manic pulsar do you have a link for the simc assassination APL converted to AMR? I saw you linked the outlaw one yesterday and I've been using that since the default rerolls anything less than 2
I don't have assassination I don't think, at least not up to date at all. Re-rolling anything less than two is the same DPS as keeping particular single buffs for outlaw - I ran a whole bunch of tests to confirm it. The update today is going to keep True Bearing just because it's not a DPS loss (although it's not a gain either).
Remember that draught also locks out other abilities so it needs to do a ton of damage on the breakdown to be any good at all
% of damage for trinkets that displace other things can be highly misleading
see cap,soul
Remember when soul cap used to explode into ring, which in turn would explode into soul cap, which would explode for 35 mil, good times..... goooood.....times.......
pulling threat from tanks was some of my proudest moments ever
yeah it was the best because that combo would pull threat twice, the first cap would go off, then they'd taunt, then the second would go off and you'd have to tank it until taunt was off cd again, good thing our class is broken AF and we have evasion
I just told the tanks I was helping them
Is the PTR mode on simcraft already updated for 7.2? Showing on the nightly build from a couple nights ago
!trinkets
has anyone done a pass on trinkets with the new changes yet?
nightblooming frond is bad
@radiant dust Hey - first, sorry that the drop downs don't work on mobile. Fixing that is high on our list (we only have 1 dev for the whole site, so his list is long). Also, @tribal scroll In case you didn't know, the reason he wanted to use mobile was because the website will talk to the client on your desktop, so you can sim from anywhere. That makes mobile AMR super handy, once we get those drop downs working, of course ๐
its bugged on simcraft. the proc rate is low and the damage bonus was cut in half on live
Did I do something wrong? http://prntscr.com/dzuoc2 Why is my haste so high?
is tirathon's betrayal simming correctly? in simc it is ~2% of my damage, but most lists have it very high
it looks like it sims similarly to a equivalent ilvl agi/crit stat stick
@split sorrel Just got Draught of Souls and it seems like in the sim its hitting for around 175k-196k when on live its hitting for around 220k - 250k
how much % of your dps is the draught?
7%
dang
Also I am pretty sure that draught isnt affected by AP but in the sim it looks like it is.
It ia
for warriors its affected by their mastery isnt it?
Lowest damage done by fel crazed Rage is 223k with AP and 223k without AP
EP affects it
Hey, I have the legi cloak that increases fok damage. how do I modify the simulation to fok once a miinute to take advantage of the proc? right now I think it only foks when there are 3 or more targets
@rustic tundra Your error bars are WAY to large
use a higher iteration simulation
50 or 100k
The latest simc build seems to have frond much lower now
About 2.5-3% of my dmg, which is much closer to live than before
Is 50-100k really the recommended iterations?
Urn slightly beats out frond for me now
Yes
10k tends to be pretty off, and if you're scaling and using pawn it could be a little brutal, switching out stat priorities and such
I personally use 100k because I have a beefy computer, 50k should be fine though
lord
Which % do you set it to
first time using, didnt set it at all
Yeah or do that @maiden cargo
You can get to within 0.25% margin of error with between 1000 and 2000 iterations in most cases. That is more than enough accuracy to make stat weights and gearing decisions.
woah, new version of simcraft put me at 267k DPS (was 497k on the version of 2 days ago)
Did your settings reset?
nah, patchwork 1 enemy, good skill, 100ping
the same i usually uses
tho i got my duskwalker footpads to 940
could it be that the sim ignore it?
same as "severe dps loss" or "same dps rate yopu had before"?
nearly exact same number as build from sunday
nightly
Can you link me a log ?
Yeah let me do one real quick
And make sure it's only hitting 1 dummy with AP on it
yeah lol it so annoying
Hmm.. must be something with skada because on logs it shows its hitting for more
but on skada it shows lowest hit was 223k
AP modifier was 38%
which is correct in logs
yeh cause all player and target modifiers apply to dos
Well AP is good then but it still hits low in simc
So you can see in the logs it was hitting for around 300k - 330k non crits but in simc it hits for much lower
@split sorrel , does sim over value the new legendary shoulders ? or are those bis for sub rogues ?
@pure sluice bis
thanks
Some fun for Assa players with boots
-
I found a minor adjustment to default APL that increases damage if player has 2 or 3 MA relics and CoF. No idea how to implement in genericly into APL or if that should be done even but here are results
http://i.imgur.com/wsA4j5x.jpg
I simply used T19_NH profile, gave it CoF and legendary boots.
Change that was made - lineactions.cds+=/vendetta,if=!talent.exsanguinate.enabled&energy.deficit>=88-!talent.venom_rush.enabled*10
now looks likeactions.cds+=/vendetta,if=!talent.exsanguinate.enabled&energy.deficit>=88-!talent.venom_rush.enabled*10&cooldown.kingsbane.remains<3 -
With that in mind I decided to investigate how good CoF with boots with regard to different fight lengths. You can find album with results below. _New is the adjusted APL as described above
http://imgur.com/a/2mylf
Looks like CoF is just good overall, no specific fights that favor other choice
sorry if its a bit offtopic. but could you tell me how to get stat weights for healer with simc?
should go to ur repsective class's discord and ask that
Oh DoS is also a 2 min CD not 1.3min like it has in Simc
I am looking at log from Jan 21st it has 1:20 cooldown there
@late willow https://www.warcraftlogs.com/reports/hQXLYrqAWHBd1Gvy#fight=9&type=casts&source=113 log from today and it looks like below 2 mins
@late willow @merry cloud
https://gyazo.com/bfd9f162263c5f813ae91f98a979133a
@merry cloud so you basically made the sim wait for Kingsbane to be almost up before using Vendetta, and 3 MA relics beat out 2 MA relics in this situation?
but that doesnt make sense vendetta would still come off cd after KB right?
does CoF really take it to under 45?
Yes it does, and consistently
so with 3MA relics + CoF
you can have vendetta line up with kingsbane all the time?
without requiring boots?
@wanton shell Yes CoF + Boots + 3MA is about a 40 sec CD on Vendetta on average
Without boots it'll be something like 1 min 12 less 15 secs, so around 1 min Vend cd
Won't align with KB
@merry cloud what was the 3rd relic you used for 2MA? Rupture or Envenom damage?
trying to get notes together for loot council. maybe you guys could help? How good are your guys 2 and 4 p set bonuses?
lorentz only simmed with the boots
his optimizations where targeted at boots + relics (+ cof)
I just used it again and it was 2 mins just to be sure
Log from this morning
Me testing it
and my log is a rogue
Last I heard it is bugged for some people
Just like ma after the hotfix
DoS used to have 2 min cd
@placid olive https://docs.google.com/spreadsheets/d/1rhg8pHjgXvtlJOvHkmDj1QDV9lIrp-b8EP-BoP_cDN0/edit#gid=0
Sheet1 Spec, NONE, 2 PC, 4 PC, 0 to 2PC, 2 to 4PC, 0 to 4PC Vanquisher, Druid Balance, 493426, 512795, 532698, 3. 9%, 3. 9%, 8. 0% Druid Feral, 462829, 463386, 477981, 0. 1%, 3. 1%, 3. 3% Mage Arcane, 477854, 496072, 497010, 3. 8%, 0. 2%, 4. 0% Mage Frost, 481100, 501372, 510723, 4. 2%, 1. ...
@final ember ty so much just saved my life
that should help convince my guild not to give our shitty ass mages the 4set before me
thats old
arms, sub, hunter bonuses were nerfed
@placid olive @wanton shell a more current sheet https://docs.google.com/spreadsheets/d/1rhg8pHjgXvtlJOvHkmDj1QDV9lIrp-b8EP-BoP_cDN0/htmlview?sle=true#gid=0
thank you
@maiden cargo you linked the exactly same sheet as 3 posts aboth
lol
rip
and i said thank you
quick question about PB proccs. has anyone actually simmed or found a way to figure out if alacrity affects the rppm of PB. and if not if in some other way it makes it more likely to procc?
@wanton shell it does effect it, all sources of haste do
i mean
BTI doesnt
memento doesnt
i thought only flat haste increases can affect rppm
and that only in some cases
if a rppm effect is increased by haste then all sources of haste will effect it
ok then why dont sin rogues go max haste with alacrity for PB proccs
like mast haste gear
the poison bomb procs arent good enough to sacrifice other stats for haste, additionally this discussion should continue in general or something as it is not a simulation style discussion
ok lets make it one. does the current APL have the ability to predict such playstyle in order to be simmed correctly?
yes in the current version of SIMC haste accurately effects the proc rate of poison bomb
because tbh you can't really be sure if it is worth to sacrifice said stats unless it's been tested right?
so you can go and set that delicious mob count to 5/7/10 and adjust yourself to have more hast
and see if it effects the outcome positively
guess i will run a full haste/mast gear sim and see the results
or negatively
why do the updated legendary sims use last patches talents for assassination? or am i missing something
any1 got draught sprint macro?
@maiden cargo you linked the same sheet as me lol, ignore the thumbnail
Why draught sprint?
thraxxis i presume
but that only affects runthrough
oh yeah true
idk then
the stopattack
lol
its for mantle for sub
to proc sprint vanish
just noticed
ahhh
clever
though I'm guessing not so much for mantly, more just to get extra vanish?
since in an ideal world, you'd have mantle going before the draught
yall know if draught works with old war
because sprint will use vanish if draught up
Thanks
!trinkets
anyone on that can help me with relic swapping and simc?
i suck a D at it
main_hand=,id=128476,bonus_id=743,relic_id=1806:1507:3336/1806:1502/3417:1527:3337,gem_id=139253/139251/136721/0 with 890 version of this http://www.wowhead.com/item=140825/felfire-pitch&bonus=1487:3514
This is a level 875 epic fel relic. It is looted from Krosus. In the Fel Relics category. Always up to date with the latest patch (7.1.5).
does anyone here have the enchant id for trained soldier for simcraft?
@finite thunder 5890
lets say you have an item in your bags that unsocketed and unenchanted, how do you sim yourself with that item but as if it were socketed/enchanted?
you add some like that ,gems=150vers,enchant=200vers by the end of item line
I ran some sims and found that for outlaw, keeping 2 buffs if one of them is broadsides is a slight dps increase over fishing for 3 or TB. Can anyone help me confirm?
Anyone have any ideas why Tirathons sims so high?
It doesn't scale with vendetta
and has never really given me much damage
Not sure if this question should go here, but here goes: Does SimC overvalue the 5 extra energy from "Stabilized Energy Pendant"? It's showing it as a 4-5k dps increase over an 880 neck of comparable stats.
My guess is yes - I noticed that too. Its extra energy on every vendetta, I can only assume the extra energy lines up with being able to use an extra GCD where you would normally pool
but thats just a guess
Oh woops. I play sub* forgot to mention. I notice in the results that the neck results have ~1.6 extra shadowstrikes. I assume this is due to being able to pool slightly more, but it seems like a rather small amount to be worth so much
It makes sense but I wouldn't stress it
base it on the stats
you can probably eye ball the upgrade
Yeah, I figured going from an 870 crit/vers neck to an 880 vers/mastery neck would be an upgrade...
yeah, no brainer
and jewellery now scales so hard
though, my 890 haste/mast ring will forever rot in my bags
literally good for nothing
has aethys did a sim for kiljadens leg trinket
If you don't see it in spreadsheet then no
Disabled in this channel - please try general ...
@manic pulsar AMR doesn't pool or have any energy rules, right?
And with deceit, it seems to cast Shuriken Storm it outside of stealth
And I really feel that opening with goremaw should be a damage loss, esp with bloodlust
Later on in the fight you also probably want to prio (non convergence/bracers) SB to be active as more outside of stealth than during
But by far the largest thing is the pooling
Does Ravenholdt have a simcraft bot that whispers you your information?
no
I dont know if this is a known issue or not, but when simming myself, with the nightly 715-02 build from earlier, to see the difference between 43 and 54 traits, I noticed that the armory import had my artifact as artifact=43:0:0:0:0:323:1:324:3:325:3:326:3:327:3:328:3:329:3:330:3:331:3:332:1:333:1:334:1:335:1:337:1:346:1:347:1:1276:1:1384:9, but when I used the artifact calculator on wowhead to make test the different traits, it showed as 43:0:0:0:0:323:3:324:3:325:3:326:3:327:3:328:3:329:3:330:3:331:3:332:1:333:1:334:1:335:1:337:1:346:1:347:1:1276:1:1384:20.
Aside from the expected difference in paragon points (9 to 20), I noticed that
it was only registering 1 of the 3 points in Toxxic Blades (323:1 vs 323:3), so envenom damage might be off by 7% in the new models. Not sure how long this has been an issue when importing a character.
ye it looks wrong with that input string(ignore the highlighted one)
Dawns can you provide your armory link?
@strong gorge AMR does have engery pooling if you want to use it. @manic pulsar might have done some tests on that in the past, I can hit him up in the morning about it.
@everyone the current armory import of simcraft is buggry for assasination.
it reads 323:3 as 323:1.
this means it only uses 1 rank in Toxic Blades instead of 3.
its highly recommendet to use the ingame addon to import your character for now.
thanks to @long ocean for reporting the bug
Edit:Armory import works in the latest nightly, however not correct in the release version
@strong gorge @reef canyon @manic pulsar It's like run_action_list and call_action_list on SimC, when stealthed it'll always go on the sub action list. While not stealthed, it'll go in the generator one only when above a give threshold. There is no use of pool_ressource in SimC Sub APL except for Shadowmeld
And for Shadowmeld, it could have been done with an action list, it's just that for 1 line it "wasn't worth"
So, AFAIK there is pooling in AMR APLs. Altough, been a while since I looked at.
How can u sim a full set of times that u dont own;
I find it really hard to believe that Nightblooming fond does so high in the sims, while in game its alot less impressive. Do you guys experience the same?
@split sorrel Does sim apl use Vendetta and Kings together or does it just run numbers as if they were used on instantly on cooldown?
is frond still siming higher than it should ?
its not simming wrong, its just really unreliable
@strong gorge Like Aethys pointed out, energy pooling can be achieved in APLs/Rotations in many different ways. You are correct that AMR does not have a built in "pool_for_next" function like SimC. Both the SimC and AMR sub rotations right now use the same structure for energy pooling. This works cleaner for sub compared to pool_for_next because multiple actions are gated behind the same energy threshold.
For comparison purposes, I created the SimC feral APL in AMR - it makes heavy use of pool_for_next. http://www.askmrrobot.com/wow/theory/rotation/b1c7a83d50744c07bfb169e81fa32b76?spec=DruidFeral&version=live&share=true In that rotation you will see "switches" with an SW icon next to them. Those are boolean variables you can use in the rotation, and I used those to effectively achieve the "pool_for_next" functionality. Not as nice as a built-in function, but the exact same functionality is created.
Also, I have tried not opening with Goremaw's Bite - I haven't seen any difference positive or negative on DPS.
Adding extra logic to shadow blades hasn't yielded me any DPS increases either, but, that of course doesn't mean there aren't optimizations to be made. If you have some more specific ideas on conditions that would yield an improvement, I'd definitely take a look. I tried adding conditions to use SB on the pull and then only if I don't have stealth - and the DPS didn't go up or down.
In those cases, I personally always change my rotations back to having the least number of conditions possible.
And, for folks interested, in our update last night I added in the Sprint-Vanish combo and also the Sprint-Draught-Vanish combo to the default rotation.
demios pointet out that using shadowblades to recharge shd is a slight gain on longer fights.
using/deleying shd didnt give me difrent results either (also not with denial and cof).
offensive sprint is a smal dps gain, its bigger with mantle and probably mroe efficient together with droughts.
if you want to optimize the apl you can try optimize the use of the finality menchanic (especialy for anticipation)or
adjust the apl to reduce the cp waste with DS.
Yeah I tested out the offensive sprint thing and was getting a slight gain even without the mantle, so I left it in the rotation
I haven't done a lot of testing with the new anticipation yet - I'll put it on my list to see if I can improve handling of finality with that talent
Using deeper strategem I never saw much of a gain worrying about it
the cp waste with 4p(+coF) is high, thats why using anticipation gives you better results however optimizing the waste of cp may end up in ds beeing better (not as a talent choice but output wise)
Not sure the offensive sprint will become a thing without mantle though. NH needs quite a lot of mobility and has its fair share of raid-wide damage, making sprint not proc vanish.
also with those optimization the question about the best talent choice reminds.
Anticipation scales the Finality buff up to 40% for evi but only 20% for NB while DS scales both up to 24% with a base increse of 5% (makes it ~30% for both with 5% boost base stuff(non finality usages))(assuming optimal usage ofc)
tbf with how NH is made, the 4pc and the NH relics I could see 2SS2finisher shadow dances being used, and Nightbane/Crit relics instead of EStab
1 ES + 2 dps relics (f.ex. gutripper) simmed the best since ages
the gain from offensive sprint in raw numbers is lower then 1% dps (without shoulders)
using sprint offensively also requires good timing and is like you sayed encounter dependent but this discussion is better fitting in #228878985210101760
Uhm.. SnD and RtB sims almost exactly the same for me, using askmrrobot
Your relics, talents, gear will affect how close SnD and RtB sim to each other
additionally the AMR APL is less refined than the simcraft one and can result in strange things occuring
Loktark - I'd challenge that... I put the SimC APL into AMR and got almost identical results
thats simply not the case
with your outlaw Simc copy APL that runs through AMR that you sent me they handle certain things differently
try it out if you'd like
especially in the category of combo point waste
well, that's just the difference in how the two simulators work
they are not identical models of the game
the results are very, very close though
between the two simulators, using the same rotation
both tools have difrent aproaches and give therefore difrent results. its up to the task and personal taste what to use
that's my philosophy as well
snd vs rtb
ill try snd in normal nh today. see how it works. easier to sustain snd than rtb
probably boring as hell tho ๐
@manic pulsar does the simc copy not perform pre fight actions?
It does. In the included rotation.
Is simc scaling it right per stack? AMR numbers are based on some tests we did. Need to test again after hotfix yesterday.
do i need to download a new simc after each hotfix?
updating with the nightly build is a good idea, yes
ty anyway ๐
what is different in it?
also, how can i see what build i'm currently using?
found it also
"git build" right?
@edgy breach can you be a bit more specific about "stack"
simc does seem to model frond stacking correctly
But it doesnt model player and/or target modifiers like ago correctly
seems like 23k lower with new build. 600k now with snd.
Is there a way to allocate more CPU usage to simulationcraft?
risking that this question was already answered: which options do you consider as useful when simulation for myth+? HelterSkelter or AddCleave, multiple targets?
@lost gale you can set set process prio and used threads in the settings menue(options - general tab)
hmm i wanna sim 940 shoulders vs 940 ravenholdt im guessing the shoulders will overtake but do anyone know how to swap id so the ilvl go up? finger2=,id=137049 is for raven for example
any idea why SIMC puts kingsbane as first source of dmg but that doesnt reflect my live dmg?
nvm wrong plot
seems like krosus cannot be hit by the "acranogolem digit"'s damage
anyone can confirm that?
didnt get a single procc in two separate 4min tries
may also be a range problem, as you stand relatively far away from the boss
so is frond good still ?
how do you sim relics? I cant seem to figure it out
@robust solar In the AMR simulator, just click on the relic slot to swap it. It's the 3rd tip on the left of this image http://i.imgur.com/r2gpuBW.jpg
Note: I believe we have to update the site for the frond proc nerf. Just fyi if you're comparing that. It will be updated tonight.
Sim Your Own Char Introduction I am Skasch, a rogue playing in Quichons (Elune-EU) and a theorycrafting lover. You can easily find me on the different WoW class Discord chans. Donโt hesitate to mention me or message me directly if you have any question! This guide is to explain step by step how ...
#4 @dark arch
thanks
actions.precombat+=/food,name=nightbourne_delicacy_platter
was is wrong with that line?
nvm found it myself ๐
hey guys, i simed my char and it says i should do around 393k but the best i can manage is 330k, where in 393k 16907 dps is from potion of old war which i dont use, do you guys know what i should do to make my real dps higher?
i know theres little chance i will reach to around 380k but making it to 340-350 would be good for starter
how do you find the item # for an item that you didn't equip yet
@onyx steeple all item id numbers can be found on wowhead
@alpine dew what are you comparing the damage to? A specific boss fight or a target dummy?
target dummy and also bosses like ursoc
on both cant manage to make more than stable 330
i would suggest posting a log of your fight against a dummy or ursoc
people are more than willing to assist you with things you may be doing wrong in your rotation
if you play outlaw you can feel free to send me a log and ill look over it to help you get your damage closer to your simulated value
hmm, im using simulationcraft (not sure if there are other types of this programme), does it simulate what my dps would be like on dummies?
if yes then i managed to make it to around stable 360k-370k
@alpine dew Do you have a recent ursoc sim, from after tuesday's patch? I have the ursoc fight in AMR, which is very similar to the patchwerk fight in SimC. What I like to do is set the fight length to match the log, along with all of the gear, relics, bloodlust timing, potions, etc. Then look at the spell usage count for each spell, and see where you might be losing damage.
this is probably dumb, but where do we modify food buffs or potions used in simcraft?
in the input string just replace them
not familiar with what the input string is
as in the simulation string that you put in?
ya i've been fumbling around looking for what to modify and haven't seen anything that looks right
Has anyone simmed Entwined Elemental Foci yet?
thanks fuu
copy everything from "profile".
paste it and make changes in the line:
actions.precombat+=/food,name=NAME_OF_THE_FOOD_YOU_WANT
and
actions.precombat+=/potion,name=NAME_OF_POTION
the name of food and potion is the same as ingame but uses underline as seperator
ya i'm seeing how to use this
does anyone knows why bracers /boots are always behind ring/boots, got the bracers today and i been trying to sim with a lot of diferent boss kill timers and they all end up with ring in first, shoulnt bracer be higher with longer fights?
i'll have to sim old war on pull and then prolonged power later
@reef canyon somehow managed to improve it, doing around 370k now ๐
Hey guys and gals if I want to test other relics for my swords with simcraft do I only need to adjust the relic_ids?
in this string
main_hand=,id=128872,bonus_id=742,relic_id=3514:1472:1813/1807:1487:3337/3414:1517:3336,gem_id=140819/139261/137365/0
Thanks
@alpine dew Glad to hear it!! That's the goal you were reaching for too, right? ๐
Is CoF modeled correctly in Simcraft? I have boots and 3x MA, and at equal ilvl I am showing a 10k dps loss equipping it over either entwined elemental foci or a crit eye of guarm.
From anecdotal testing up to 500m damage on a target dummy though it's definitely an increase to equip it over either.
Is there a change we need to do for draught of souls simming?
Well, you need to use it. If you just change from a passive trinket to DoS without regenerating an APL it'll likely not be using it.
Also there's probably quite a lot of optimization on when you use it you can play around with.
I added a quick line for DoS as Subtlety, but it's likely it need improvement on all 3 APL for the On Use usage since this one need to be used On CD.
did anyone sim a profile with CoF + 3 MA relics since the kingsbane "bug" was found? i guess its worth to wait for kingsbane or switch one MA relic for rupture/envenom. If not, how can i change the vendetta line to sync it with kingsbane?
Also, will have a look at Sprint / DoS someday.
@manic pulsar Thanks. I have to admit I only looked at what you call "Rotation" in the result and compared it to the log generated. And it looked like quite many cases where it uses SD, ends up at ~80 energy after being used. And then being <20 energy starved for using the 4th (and in some cases even 3rd) Shadow Strike
Since both are 3s channeled on SimC, I'm not sure if I should modifiy the sprint offensive according to this, or hardcoding some thing to trigger it;
Question for the peanut gallery here: does Goremaw's Bite give energy back continuously, or discretely?
I.e. 30 energy in six seconds, or is it 5 energy at 1s, 5 energy at 2s, etc
id imagine discretely because game mechanics have a tick rate when they are over time effects
what that interval is, not sure
I don't think it's working right, but I want to test a bit more.
I'm playing around with updating the sub APL for anticipation, because I think it might be optimizable pretty aggressively
Closing the loop: it's five energy each second, for six seconds.
All
@strong rune what do u want to change for sub anticipation APL?
putting in quite a few conditionals, tweaking the sprint stuff, extra micro-complexity around goremaw's bite when CP are coming
at a high level, imo, sub is all about correct combo point management
because in a sense, using your finishers gives energy back, so you can directly turn CP efficiency into energy efficiency
taking advantage of the anticipation bug too
i.e. something like an actual cycle becomes meaningful
9-10 CP eviscerates followed by 4-5 CP eviscerates is the equivalent of an extra 20% damage on both, on average
Is the frond simming correctly or abit higher than it should ? It only does 1-2% of My dmg everyone. I can send logs etc if that helps
frond seems to be high variance right now
Ye is it still worth having or shall i swap it ?
unclear; depends on how highly you value consistency
why is using 9-10cp folowed by 4-5 cp an equivalent of 20% extra damage on both that doesent seems right
10CP puts your into 40% finality on the next one
Well since im mythic progressing now i think about swapping it for 880 Angerboda or urn
Its 890
And already got socketed cof
ah okay you mean generaly not in relation to using it on 5 cp
yeah, fuu.
Its sims like 25k higher than urn but not sure tbh
interesting thought experiment on mantle-openers as sub
assuming MoS and Anticipation
SS, SS, SS, Goremaw's bite. If you got an ES proc, immediately evis, evis
if you did not get an es proc, just pool for a very short time
you can consistently fit both eviscerates into 100% crit window
no rng
then you just dance as per usual
doesn't require boots or anything
it delays nightblade by a couple of GCDs over the standard stuff, but I'd imagine two super-buffed eviscerates might be worth more
shadowblades-ss(3)-ss(3)-gorens(3)-finish(9-10)-finish(4-5) would probably be better
hmm, because they're both buffed with MOS?
that gets them both in during the 5s window
becouse both finisher fit into the mantle buff
both finishers do already fit into the mantle buff, though ๐
even without blades
want a gif?
wasnt it 5 secs on opener?
5s after the first GCD spell
so it ends up being 6
as in, you can fit 6 GCDs with 100% crit
okay that was my mistake then^^
that being said, you might not be wrong
because they do NOT fit into the 10% MOS window
Do u get subterfuge from Sprint vanish trait?
Would it be Worth it to presprint?
To get 3 more extra seconds for shoulders buff
i'd conjecture that one is not worth it, napkin-math
you'd much rather have that sprint later in the fight
well, i don't know about much rather but this is feelycraft without data, sorry
just thinking about abusing 100% crit
anticipation definitely makes things more interesting
whole thing feels like a giant puzzle
the difrence to DS on the opener shouldnt be high
the finisher damage is 470% for DS and 480% for anticipation (from my napkin math)(during opener with mantle)
does anyone know if the trinket buffs are implemented in a simc nightly build yet
after a bit of optimization tonight i think shoulders might be simming low
on sub
i'm simming at 555k and was fairly consistently hitting 585k on krosus using sprint on cd and with no add padding
totally possible that was just variance in ES procs tho
apl is not perfectly optimized for mantle yet so better results then simming arent wrong
does anyone knows why bracers /boots are always behind ring/boots, got the bracers today and i been trying to sim with a lot of diferent boss kill timers and they all end up with ring in first, shoulnt bracer be higher with longer fights?
insignia is a little bit off on SimC with Vendetta
and this little bit thing makes it looks like better
and tbh, in progress, bracers are better even if insignia was ahead (except in cleave situation)
I just put together a weakaura to track Mantle Sprint eligibility
i.e. if you sprint and something ruins your sprint
it will show hat
would anyone else want this, or is it non-useful?
(if you take damage)
yes @strong rune
yea that's what i though, i just wanted to try and get a boss timer so i can use as reference so i can decide what bosses should i go with ring or bracers
post it in theorycrafting to spread it imo since it's a hot subject atm
@odd sedge - I also play outlaw ๐
was realy disapointed with frond, first was happy to get a 900+socket one, but then after seeing his damage ๐
I think sub is the most interesting spec with the most potential for exploration still.
There are minor things that can be done for assassination, for sure, but other folks are more competent and working on those better than I could.
Frond is really random. Sometimes 2% and other times 10%
What i was thinking is
If u look at sim ur haste value goes close to 10 atleast for me
So frond might be alot better if u actually starts to get some haste ? Thats just a thought , thoughts @split sorrel
The problem with frond atm might be that we skip haste entierly as assa so get some haste if you have might higher the frond alot , havnt Done that math tho
Do we have any info on DoS for sub? Like coupling it with sprint and shit, how does it fair against other trinkets
BiS
sooo I'm guessing CoF and DoS? Also I'm guessing depending on the situation I shoudn't always wait with sprint vanish to pair them since there are times where we can take 3secs of downtime esp with anticipation
anyone know why i sim more with simc-addon import than armory import? are the profiles different or what? I have double checked that both have same gear and same prepot/food. Is the apl different somehow?
look at the pinned post from @regal agate
@modern aurora thanks. i shouldve checked that before asking
are you required to put an apl in your simulate profile? i thought it just pulls the latest one automatically when nothing is provided.
and you only need it to change things like food
@cerulean junco you can change everything if you want but mostly for testing purposes like trying to find a better opener for your character setup
So I got nightblooming front for assassination yesterday, and i'm not really hyped about the damage I get from it. Why does it Sim so high?
but the profile createt by !simc addon is sufficient, right?
@split sorrel from my last simulation MP+Alacrity is the best build for single and multiple targets fights... what do you think?
update simc
i have 715 02
if you want accurate you use nightly builds
if you want accurate, one 1 day because I'm updating several things
how much mastery do you have?
me?
yup
129% up to 150% changing rings
with 150% which build sims higher?
I am curious to which build will sim higher with really high amounts of mastery
today i try simming MP+Alacrity for 3 target and is clearly better fore aoe
than i do for 1 target
and is still better than MP+Vigor+Ago
that's a surprise for me
for now, for any 2+ enemy fight alacrity seems better, I do believe with the set ap scales much better for single target
I don't know how to sim properly the scaling in dmg for each build tho
maybe the mastery + haste combo can end up beating ap? no idea
altho alacrity also loses more dmg from doing mechanics
@split sorrel do you know of anyway to test openers in simc, perhaps a cast sequence and very short fight length?
you can yes
ok... maybe is better for me that i have legg boots? :V
more haste more energy
faster reduction of vendetta cd
are the 2p and 4p bonus included in simming?
no i haveo only 1p
it was a general question :>
@split sorrel sorry for bugging you do we have any recent results on relic choice with 2/4p for sin? cause ive seen stuff all over the place. even putting master alchemist ahead of MA
Master Alchemist isn't a bad relic.
but is it ahead of vendetta ones?
!relics
wrong channel
sorry
i thought i was talking in general
gonna take it there
the thing is it depends on fights, are your fights long where you can multiple cast vendeta then MA > MAL
and legendary boots
i see several rogues just sticking with 3 MA relics irrespectively just bcos of a short cooldown
approx ~~ 1 min cooldown with CoF / (And godbless legendary boots) is a good dps % increase
i have the boots
then 3 MA hands down
and pretty decently leveled relics
also got priority on next cof
would it still be over GW?
Guys , What do you think about Draught of Souls
From guldan?
For assassination @split sorrel has an updated list of good trinkets
!trinkets
!trinket
Assassination Outlaw Subtlety Sim Credit goes to Aethys#5243 on discord. His site http://simc.aethys.eu/ contains many more simulations Note: Nightblooming Frond is currently simming higher than it should be. Draught and CoF are most likely better. Assassination (adsbygoogle = window.adsbygoogle || []).push({}); Created with Highcharts 4.2.3
i mean
nightblooming cant be right
unless it uses the ptr version
ah tehre is a note
ye
also another long standing question. do the bracers make BTI better or are they mutually exclusive? again sorry for asking here but noone else seems to give answers this early in the day
if I wanted to sim a new relic, specifically this one http://www.wowdb.com/items/140816-fingernail-of-the-fel-brute?bonusIDs=3516 what would I need to change on my relic line in simc, the way its set up is a bit confusing main_hand=,id=128870,bonus_id=741,relic_id=1806:1502/1806:1502/3454:1472,gem_id=139251/139261/143524/0
http://www.wowdb.com/items/139261-tuft-of-ironfur?bonusIDs=1806 thats the relic i'd be replacing if that helps
I think that you change the gem_id fro 139261 to 140816, then change the bonus ids from 1806:1502 to 1487:3516
for future reference because i'd like to be able to figure it out for others, wheres the 1487 come from?
im simming by far highest with frond and entwined 890
i can see the 3516 as the bonus id
even more then covergance or urn
ppl saying entwined is shit tho?
feel like it procs often..
Melko, it might work without the 1487, I'm not totally sure.
I got the IDs from the UI on AMR. We show the gem id and all associated bonus ids on the tooltip
I've been poking around trying to figure it out, I think it should be 1487:3514 for an 890 normal warforged, but just 3516 for the baseline hc
you can set the ilvl with relic_ilevel= if you prefer
which part would you replace to do that?
main_hand=,id=128870,bonus_id=741,relic_ilevel=880/880/880/0,gem_id=139251/139261/143524/0
for example
gem_id is the itemID of the relic
ilevel is its ilevel
got it, thanks for the help
also, use a nightly
yea I got the most recent one just a few minutes ago
actually dunno as Sub if the Nightblooming Frond hc version is fine (like in the simcraft resoults) or if it's just a simulationcraft bug that shows an higther dmg despite of the recent buffs to the other trinkets
but i think it's fine going with a Nightblooming Frond 890 and an angerboda 890
with the most recent nightly build frond is simming lower than cof and drought, so you might want to get that if you havent
i'm not sure, i don't play sub
I had someone do some testing of frond for me in-game, and the damage on it is higher than the spell data suggests. I'm not sure if SimC is accounting for that or not. I have a multiplier on it in the AMR simulator for now.
but the value of frond seems to have been adjusted, so results for sub may be different also
The tests showed it about 36% higher than the average value the spell data indicates.
at least, that is what is matching logs right now
If anyone else wants to give me more tests, I'll try to confirm the damage. I had a demon hunter test it for me.
i've been using it since last wednesday and the variance is fairly massive
i have an 880 version and the damage it deals varies between 2% and 10%
the piece of information i would need to verify it
is if you hit a boss dummy, how much damage does the first hit do
best to test it naked, to avoid any multipliers
thats going to be hard to test, stopping your autoattack after a single hit is going to require some pretty good reactions
you don't need to stop it
just tell me how much damage the first hit did
after that it just scales up 50% per hit
can you take weapon off?
do it bare knuckle
do it bare knuckle
so you don't get artifact damage multiplier
it doesn't proc
1rppm ftw
are you sub?
sin
ah, ok, then yeah that adds up to what I have right now
in the simulator
seems like the first hit is about 36% higher than what you'd get out of the item data
so simcraft is actually undervalueing it?
I don't know for sure - would have to ask @split sorrel if they have adjusted it or not to account for this
the issue i have with it is that its so erratic, the variance from fight to fight is massive
yeah, low rppm trinkets are like that
lots of people don't like them for that reason
only thing we do besides using spell data is the 50% malus per stack
I assume this 36% dmg increase is from an hotfix ?
Well, rather stealth hotfix ?
I don't know - I haven't seen it actually show up in the game data
I've just had to reverse it from logs
and tests
cool, let me know if you see anything different than what I'm seeing
what item level?
so, how is SimC calculating the damage of the proc?
the first one ?
since when I look in the spell data SimC has, I see an average of 2.12
and then item level 895 is a budget of 3938
so that is 8348 base
and this player has 14.5% damage from artifact, so that puts it at 9558
where does the rest of the damage come from?
I didn't coded it, I just hotfixed the 50% thing on each stack
might be some things from shaman
Hi outlaw here can someone tell me how come the sim says like 100k more dps than i can possibly do? like 470 sim i do 370 (With a lot of luck on rolls)
got a pastebin http://pastebin.com/DUi3DpyX
a link to a log of you playing outlaw will be far more helpful than a link to your simulation
in all likelyhood your making a few mistakes that are costing you dps
or your compared a single target patchwerk sim to a fight that doesnt allow for 100% uptime
im am comparing with me on dummy (no BL though but it cant be so much)
ill get the logs
should i get other boss maybe? should have all nighthold
yo @round canyon
The bug with mastery negative scaling was fixed. It can be traced back to blade flurry's implementation deferring the scheduling of damage to later
the literaly most recent commit solves this problem
navv did some good digging
Anyways, I used the exact profile I had been using to test the bug and this confirms to fix it
Yes I wasn't able to fully understand what was going on, but the addition of RT into BF made that schedule issue. Navv helped us ๐
are there any sims for race choices for Sin rogues
the impact here is actually pretty substantial - @split sorrel, we might need to re-run all the multiple target stuff for outlaw
before the fix:
http://pastebin.com/cth8iVjw
after the fix:
http://pastebin.com/dEffETsu
take a look at the magnitude of that fix
on the order of 16% for that particular configuration, in the low mastery
(the sane test)
Have anyone simmed all the new trinkets in Nighthold? cant find any information whats BiS anytmore
by the way, i think its worth rerunning sims in NH gear
a lot of talents and such are affected by it
Does anyone know if CoF is more valuable for assassination with or without legendary boots?
in any case, i'd make the claim right now that Blade Flurry's simcraft model has been horribly broken in game for ... a long time.. and we need to reevalute all the outlaw multiple target stuff
in particular, simc would probably have been devaluing mastery for a super long time
from that error i found?
yeah!
short answer: suppose you had a big-ass run through pumped into blade flurry
as your mastery went up, it looks like it was replacing that run through with main gauche damage
which is way lower
and deferring the run through execute to "single target" or whatever (not into BF)
so that's the negative mastery scaling you saw: as you get more MG procs, it was erroneously pumping less RT damge into BF
makes sense
so this is a pretty big bug, imo
and merits revisiting a lot of stat-weight assumptions re: outlaw
for what it's worth, the AMR sim has been valuing mastery a lot higher for outlaw for a LONG time
and I could never figure out why - this is probably related, or the root cause
yeah but this related issue is only effecting BF though right?
near as i could tell from all the stuff i looked at AMR had a higher mastery value because of the differences in the APL
also possible.
of course nowadays both programs(simc/amr) seem to be putting all 4 of my secondary stats at being so close in value that it doesnt matter
I just tested that a couple days ago - the APLs do pretty much the same thing in AMR. I didn't see any major difference - damage was almost identical. (and yes, outlaw secondaries have always shown as very close in all my tests)
Yep, I saw that.
this is a pretty degenerate case involving, like, ten target cleave or whatever
loktark sent us a profile that displayed negative mastery scaling
as in, add 8000 mastery, you do 5% less dps
i mean the mastery bug in simc is noticable at even 3 target cleave
yeah.
where it has mastery at half the value of my other stats
when it absolutely should have an increased value with more targets
anyways, it's not in the nightly yet, but if you build from source you can get the fix pulled in
also i ended up doing the adding 8k mastery thing because at 7 target cleave i was getting a negative scaling value for mastery
i was curious if it actually was providing negative damage
whelp. lots of bugs waiting to be discovered, I guess
thanks @manic pulsar - good stuff!
the constant battle of simulation, heh
@manic pulsar for what its worth im quite impressed with the latest amr stuff, the only tiny issues with it i have are use effects from say gnawed thumb ring
find the bugs, get em fixed
yeah frozen - you can add those explicitly to the rotations if you would like
its really handy to efficiently find a good combo of gear to work with without messign with a ton of statweight balancing
to optimize
i've slowly been going through and adding potion optimizations
and draught use
this was the only thing i found a bit weird
0 vers goal for machine learning weights
it's plausible that that's correct
so, those "goals" are a bit misleading -
there are likely many configurations with some vers that are near-optimal.
(the directionality, at least)
we're working on a new visualization for the machine learning
ooh, neat!
the "stat goals" were kind of a failure, heh
i think the recommendations it gives are overall very good
but the goals are kind of hard to interpret
i think the premise is tough to communicate but the architecture is totally 100% sound
yeah, the underlying function is actually very good, but the goals are too narrow - they don't describe to you what the function will try to do
i.e. estimate gradient of local surface as multiple linear regression against many points in neighborhood
we came up with a visualization that is 2D, which we believe can represent this multi-dimensional problem
that's awesome. you can always do something like "lagrange multipliers" vis or whatever
where you visualize the isocline DPS surface
as contours on a 2d plot
and then you project that
for all the pairs of dimensions
im curious how the "human delay" setting actually works in amr
that i can explain
demios is starting to use words that only he and yellowfive can talk bout
"human delay" is basically reaction time
how the human delay factors in is best explained by example
does machine learning end up producing something similar to those 4d plots aethys had?
how does the reaction time impact things though, like one were to react to getting an opportunity proc before the gcd is over
does it properly queue up the ability to cast at the end of the gcd?
so, here is how AMR works, in a nutshell
lets say you use an ability at time 1.0
your next ability you want to happen at time 2.0
in-game, we know that you actually have to decide what to do and press the button some time before 2.0
or else it won't chain perfectly
this is the "spell queue"
our testing shows that the game allows you to "press the button" so to speak up to 300ms or so early
"early" being
before it is ready to be cast or before the gcd is up
right
we default this to 250
because good players will usually wait a little longer
so, then
you have to decide what to do. press the button, overcome lag, and then have it happen
all before time 2.0
so the "anticipation time"
decides when you, the player, decide what to do
we give this a range by default
of i think 283-315 ms
er 285-325
seems totally defensible, and in line with human reaction data
so, at time 1.7-ish, you will decide what to do
this is a really good insight, thanks for the expl!
then, you have to press the button
(looking at notes for one sec)
so, what the simulator does is it creates a "decision cutoff" at end of current GCD - anticipation time
so in this example, lets say 1.7
any piece of information that you gain between time 1.7 and 2.0 is invisible when picking the next action
The evaluation of what action can be taken occurs at Decision Cutoff + Human Delay
so, if the action is usable at that time, it lets you do it
interesting
the lag gets added on to this as well. But, lag only really comes into effect if Human Delay + Lag is greater than anticipation time - which mirrors what we see in-game during lag spikes
you end up with gaps between abilities
Human Delay also comes into effect when you are in "reaction mode" - that is, you are waiting for an action to become available (waiting for energy, react to a proc, etc.)
you can't act immediately when you gain a buff, or hit 50 energy, or whatever
there is a delay
you can use .react in SimC for gaining buffs
to mirror this
but you have to put it in APL - it's not built in to the "guts"
for melee specs - all of this has very minor impact
for caster specs, it's a much bigger deal
but, you will notice that SimC, especially for sub rogues, will always get a few more shadowstrikes